Heterosis, or hybrid vigor, describes the increased yield and robustness of hybrid plants relative to their parents and is a cornerstone of modern crop breeding1. Beyond biparental heterosis, autopolyploid progressive heterosis (APH) is observed in maize, potato and alfalfa when genomic segments from four distinct grandparents are combined, leading to additional heterotic effects2. APH has yet to be fully exploited in commercial breeding because meiosis reassorts genotypes and genetically uniform seeds that benefit from APH cannot be produced. The ‘mitosis instead of meiosis’ (MiMe) system previously established in Arabidopsis and rice leads to clonal, unreduced gametes3,4,5,6,7, but has yet to be established in a dicot crop or tested in the engineering of polyploid genomes by design. Here, we established polyploid genome design in tomato to allow the controlled combination of four predefined genome haplotypes through the hybridization of clonal gametes produced by two distinct hybrid parents.

We set out to establish a MiMe system in tomato to produce clonal gametes in a controlled manner. Building on fundamental insights from tomato meiotic mutants (Supplementary Note 1), we found that a functional MiMe system could be established in inbred tomato through mutation of SlSPO11-1, SlREC8 and SlTAM (Fig. 1a−c, Extended Data Figs. 1 and 2, Supplementary Figs. 116 and Supplementary Tables 14). We implemented the MiMe system in three hybrid tomato genotypes, including the Moneyberg-TMV Micro-Tom (MbTMV-MT) model hybrid, the date-tomato commercial hybrid ‘Funtelle’ and the truss tomato commercial hybrid ‘Maxeza’ (Fig. 1a−c). We identified two independent MbTMV-MT, three independent Funtelle and three independent Maxeza lines with biallelic mutations in SlSPO11-1, SlREC8 and SlTAM (Supplementary Table 4). We focused on one putative hybrid MiMe line per hybrid and found that all were capable of producing unreduced pollen (Fig. 1c). We prepared chromosome spreads from these three hybrid MiMe lines for cytological analysis of meiosis (Fig. 1d and Supplementary Fig. 17). In wild-type meiosis, homologous chromosomes synapse and form 12 highly condensed recombining bivalents, which is followed by two rounds of segregation to generate tetrads (Fig. 1d). In contrast, the hybrid MiMe mutants went through a mitotic-like cell division in which 24 univalents were identifiable at diakinesis, followed by the production of dyads with a single round of segregation (Fig. 1d and Supplementary Fig. 17). Compared with the MbTMV-MT fruits (20.22 ± 2.04 g), the MbTMV-MT MiMe mutants produced smaller fruits (11.31 ± 0.74 g) that contained fewer seeds (Fig. 1e,f). However, these seeds were larger than wild-type seeds and gave rise to tetraploid offspring at high penetrance (93%) (Fig. 1g,h and Supplementary Fig. 18). We sequenced tetraploid offspring from two MbTMV-MTMiMe parents (six offspring from each) and controls and used polymorphic genetic markers between the parental genomes (MbTMV and MT) to infer crossover recombination (Supplementary Fig. 19). As expected, F1 plants were heterozygous across the genome (0.5 allele frequency), whereas F2 plants showed divergence to homozygous states (0 and 1 allele frequency), indicating that crossovers had occurred (Fig. 1i). In contrast, tetraploid MiMe offspring (0.5 allele frequency) demonstrated a pattern similar to that of the F1 hybrid controls without crossovers (Fig. 1i). In some MiMe offspring (3/12), we observed local deviations from the 0.5 allele frequency (Fig. 1i) and normalized read coverage, indicating chromosome instability (Supplementary Fig. 20). In one of the offspring (MbTMV-MTMiMe-A_6), we validated the partial loss of one of the MbTMV copies of chromosome 9 (Fig. 1i). These chromosome truncations may arise due to SPO11-1-independent DNA double-strand breaks (for example, arising from DNA replication stress or environmental DNA damage) that could not be repaired by homologous recombination due to the absence of homologous chromosome pairing and disturbed sister chromatid cohesion. Next, we explored the phenotypic behavior of hybrid MiMe offspring and control F1 and F2 plants, focusing on plant height, fruit and seed development and leaf morphology (Fig. 1j,k and Supplementary Fig. 21). We found highly diverse phenotypes among the MbTMV-MT F2 offspring, whereas highly consistent phenotypes were observed among the F1 hybrid and MiMe tetraploid offspring (Fig. 1j,k and Supplementary Fig. 21). These findings illustrate that unreduced, clonal gametes can be produced in hybrid tomato plants containing mutations in the MiMe genes and that their tetraploid offspring maintain genome-wide heterozygosity and plant characters.

Fig. 1: Clonal gametes by MiMe lead to the inheritance of genome-wide heterozygosity and plant phenotypes in tomato.
figure 1

a, Schematic of establishing MiMe in four tomato genotypes. Created with b, Alexander staining of pollen from Micro-Tom (n = 32) and spo11-1 rec8 tam (n = 83). Scale bars, 100 µm. c, Single-pollen diameter in wild type and MiMe plants. Pollen grains below the red dashed line are deemed as nonviable pollen grains. P values were calculated using the Wilcoxon rank-sum test: ****P < 0.0001. See the ‘Statistics and reproducibility’ section in the Methods for more information. d, Chromosome behavior of male meiocytes in the wild type and MbTMV-MTMiMe-A. Wild type: pachytene (n = 53), diakinesis (n = 46), anaphase I (n = 35), dyad (n = 28), tetrad (n = 48); MbTMV-MTMiMe-A: diakinesis (n = 47), anaphase I (n = 56), dyad (n = 57). Scale bars, 10 µm. e, Transverse anatomical view of MbTMV-MT F1 and hybrid MbTMV-MTMiMe-A fruits. Scale bar, 1 cm. f, Seed number per fruit in the wild type (n = 74) and MbTMV-MTMiMe-A (n = 75). The median is shown by a solid black line and quartiles are shown by dashed black lines. The P value was calculated using an unpaired two-tailed t-test: ****P < 0.0001, with an exact P value of P = 4.99 × 10–5. g, Flow cytometry analysis of diploid parent MbTMV-MTMiMe-A (black) and tetraploid MbTMV-MTMiMe-A (red) offspring. h, The ploidy level of MbTMV-MTMiMe-A offspring estimated by flow cytometry of leaf nuclei. Tetraploid plants (n = 77) were validated and six plants were potentially aneuploid. i, Whole-genome sequencing and allele frequency analysis of an MbTMV-MT F1 plant, two MbTMV-MT F2 plants and two tetraploid MbTMV-MTMiMe-A offspring. The allele frequency distribution between MbTMV (0 on the y axis) and Micro-Tom (1 on the y axis) is shown. Meiotic crossover positions are approximated with black dots. j, Time series of plant height of MbTMV-MT F1 (green), MbTMV-MT F2 (blue) and two MbTMV-MTMiMe offspring populations (red and brown). k, Fruit morphology of MbTMV-MT F1, MbTMV-MTMiMe-A offspring, MbTMV-MTMiMe-B offspring and two different MbTMV-MT F2 plants. Scale bar, 2 cm.

Next, we harnessed the MiMe system for polyploid genome design to generate plants that contained the complete genetic repertoire of their four inbred grandparents. According to classical nomenclature, such plants could be referred to as ‘nonrecombinant, double-cross hybrids’ but, for simplicity, we refer to them as ‘4-haplotype’ (4-Hap) plants2,8. To generate 4-Hap plants, we designed two sets of crosses between hybrid MiMe plants (MbTMV-MTMiMe MaxezaMiMe; MbTMV-MTMiMe FuntelleMiMe) (Fig. 2a and Supplementary Table 8). We used a platinum-grade genome assembly of the inbred grandparent MbTMV9, generated a platinum-grade assembly of Micro-Tom and developed haplotype-resolved assemblies of the hybrid parental lines (Funtelle and Maxeza) to identify unique single-nucleotide polymorphisms (SNPs) for each haplotype in each cross (Supplementary Note 2, Supplementary Figs. 2227 and Supplementary Tables 5 and 6). Next, we aimed to fully characterize the distinct haplotypes in a set of 18 putative 4-Hap plants (13 from MbTMV-MT-MaxezaMiMe and 5 from MbTMV-MT-FuntelleMiMe) using whole-genome sequencing. Using the haplotype-specific SNP markers, the presence of all four parental genomes in each 4-Hap plant was validated (Fig. 2b,c). In addition, we found that each 4-Hap plant inherited mutations in the MiMe target genes, which confirmed genetic inheritance from both parents (Supplementary Table 4). Furthermore, cytological analysis of a subset of lines demonstrated that 4-Hap plants had the expected 48 chromosomes (Fig. 2d and Supplementary Fig. 28). By examining allele frequency, we confirmed equal dosage from both parents and tetraploids in 16 of 18 4-Hap plants analyzed (Supplementary Figs. 3135). Chromosome truncation was observed in MbTMV-MT-MaxezaMiMe-9 (Supplementary Fig. 30). We predicted the agronomically relevant gene dosage in 4-Hap plants based on parental genome sequences and subsequently counted the allele frequency in the 4-Hap plants themselves (Fig. 2e). This revealed that genotypes were inherited as expected if gametes were clonal (Fig. 2e). Synteny analysis of the Funtelle haplotypes and annotation with a Meloidogyne incognita (Mi-1) resistance gene marker confirmed introgession in the structurally divergent Funtelle haplotype 1 (Fig. 2f and Supplementary Fig. 32), which was further elaborated by elevated numbers of Solanum peruvianum (Mi-1 donor) genes along that haplotype (Fig. 2g). The genotyping analysis showed that MbTMV-MT-MaxezaMiMe plants had more copies of the Tm-22 haplotype than MbTMV-MT-FuntelleMiMe plants (Fig. 2e and Supplementary Fig. 32), as predicted from the parental genome sequences and the distribution of S. peruvianum (Tm-22 donor) genes (Fig. 2g and Supplementary Fig. 33). Remarkably, phenotyping tests of 4-Hap plants demonstrated normal vegetative growth and the production of well-organized inflorescences that harbor seedless fruits, as well as higher chlorophyll content in mature leaves (Fig. 2h,i and Supplementary Figs. 30, 31 and 36). In summary, we found that 4-Hap plants contain four genomes directly descended from their four inbred grandparents and allow a unique combination of plant characters.

Fig. 2: Precise engineering of tetraploid plants with four nonrecombined haplotypes via polyploid genome design.
figure 2

a, Schematic of generation of tetraploid 4-Hap plants that contain the complete genetic repertoire of their four inbred grandparents. Created with b,c, Presence of haplotype-specific markers in 13 MbTMV-MTMiMe MaxezaMiMe offspring, 2 Maxeza F1 plants, 5 MbTMV-MTMiMe FuntelleMiMe offspring, 2 Funtelle F1 plants and 3 MbTMV-MT F1 plants. Colors represent each haplotype tested and the size of the circle represents the percentage of markers found (a circle completely filling the square denotes that 100% of markers were found). d, Chromosome spreads of male meiocytes from a tetraploid MbTMV-MTMiMe FuntelleMiMe offspring 4-Hap plant at the diakinesis stage (n = 36). Scale bar, 10 µm. e, The expected gene dosage (histogram) and whole-genome sequencing-based genotyping (box-and-whisker plot; solid black line is the median, boxes show quartiles and whiskers show values within 1.5 the interquartile range above and below the quartiles) of MbTMV-MT-FuntelleMiMe (n = 5) and MbTMV-MT-MaxezaMiMe (n = 13) 4-Hap plants. The genes tested encode tobacco mosaic virus resistance (Tm-22, Solyc09g018220), M. incognita resistance (Mi, Solyc06g008720), self-pruning (SP, Solyc06g074350), dwarfism (D, Solyc02g089160) and Fusarium wilt resistance (I, Solyc11g011180). f, Genomic rearrangements on chromosome 6 at the Mi-1 resistance locus. The haplotypes are depicted as horizontal lines and are (from top to bottom) MbTMV, Micro-Tom and Funtelle-2 and Funtelle-1 haplotypes. g, Frequency of genes derived from S. peruvianum (donor of Tm-22 and Mi-1) per genome haplotype. h, Images of a tetraploid MbTMV-MTMiMe FuntelleMiMe offspring 4-Hap plant. Whole-plant morphology (left), structure of a branch with ripening fruits (middle), fully ripened fruits (top right), harvested fruits (middle right) and cut fruits (bottom right) are shown. Scale bars, 2.5 cm. i, Left, single-fruit weights of MbTMV-MT F1 (n = 40), Funtelle F1 (n = 34) and four tetraploid MbTMV-MT-FuntelleMiMe plants (1, n = 41; 2, n = 53; 3, n = 46; 5, n = 52). Right, single-fruit weights of MbTMV-MT F1 (n = 40), Maxeza F1 (n = 17) and 12 tetraploid MbTMV-MT-MaxezaMiMe plants (1, n = 27; 2, n = 31; 3, n = 36; 4, n = 18; 5, n = 35; 6, n = 28; 8, n = 11; 9, n = 23; 10, n = 15; 11, n = 33; 12, n = 30; 13, n = 25). The same MbTMV-MT F1 data are used twice in this panel for comparison. We used ordinary one-way ANOVA followed by Šídák’s multiple-comparison test: P < 0.05, ‘NS’ (not significant); *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. See the ‘Statistics and reproducibility’ section (Methods) for more information.

Polyploid genome design has the potential to control genetic heterozygosity in polyploids, thereby allowing APH to be fully exploited in agriculture. In this report, we demonstrate that clonal gamete production in hybrid crop genotypes allows precise polyploid genome engineering; however, the exploitation of APH will involve further steps. This will require the development of four-way heterotic groups, which could be driven by using genomic selection to identify higher-order combining abilities between grandparental lines10. Tetraploidy doubles the length of the genetic map of a diploid crop, meaning that breeders could incorporate more unique characteristics in elite lines that were previously abandoned because of polygenic inheritance or prohibitive linkage drag. For example, our blueprint could facilitate the introgression of one or multiple complete ‘wild’ genomes into cultivated crops to facilitate abiotic and biotic stress resistance encoded by several unlinked genes11,12,13. It has not escaped our attention that polyploid genome design has major implications for hybrid potato breeding, as it provides the flexibility to perform genetic improvement at the diploid level and then harness heterosis at the polyploid level on the farm14,15,16. Taking a wider perspective, polyploid genome design could be used for the clonal transfer of genomes from diploid wild materials into current polyploid crops (for example, strawberry) and the generation of highly heterozygous seedless triploid varieties (for example, banana). Hence, we propose that polyploid genome design can facilitate the controlled increase of genetic diversity in crops and open up completely novel breeding schemes.


Plant growth and materials

Tomato plants were grown under long-day conditions (16 h light and 8 h dark) in Bronson chambers, Percival chambers and greenhouses at the Max Planck Institute for Plant Breeding Research. Tomato hybridization was accomplished by manual emasculation and pollination. For seed origin information, please refer to the Acknowledgements.

Seed processing and germination

Seeds were collected from ripe tomato fruits and the pulp seed mixture was cleaned in a 1:1 (v/v) mix with 2% HCl solution for 30 min, followed by washing with fully desalinated water and drying overnight at room temperature. Thereafter, the harvested seeds were dried at room temperature for several weeks and stored at 4 °C. Seed germination was performed in vitro. Seeds were incubated in 1.5 ml sterile Milli-Q (Millipore) for 2 h. Subsequently, seeds were incubated in 1.5 ml saturated Na3PO4 buffer for 20 min, washed in Milli-Q three times and then incubated in 1.5 ml of 2.7% sodium hypochlorite (NaClO) for 20 min and washed three times using Milli-Q. Next, we transferred the seeds to 0.8% agar at 25 °C and then transferred the seedlings with roots into half-strength Murashige and Skoog (MS) solid medium in a culture room or soil in the greenhouse.

Seed imaging

Seeds from wild-type Micro-Tom, Sltam, MbTMV-MT F1 and MbTMV-MTMiMe-A mutants were imaged using a Leica M205 FA digital stereomicroscope (Leica Microsystems). Thereafter, Leica LAS X software was used to analyze the images. For quantitative seed size analysis, seed images were processed using the ‘threshold’ feature of ImageJ ( Seed size area was measured using the ‘analyze particles’ feature, with a lower limit of ‘1-Infinity mm2’ to exclude any nonseed material. The final data were analyzed using Microsoft Excel and GraphPad Prism 9 software.

CRISPR−Cas9 vector construction

The CRISPR−Cas9 system vector pDIRECT_22C17 (containing a 35S::AtCAS9 cassette and in planta kanamycin resistance) was acquired from Addgene (plasmid no. 91135; and modified to knock out SPO11-1, REC8, TAM and OSD1 (ref. 17). Specific single guide RNAs (sgRNAs) targeting each gene were designed using CRISPR-P v2.0 ( and were selected to have a low rate of off-target mutagenesis in tomato18. The specificity of the designed sgRNAs was checked against various tomato genome assemblies, including BGV006775, BGV006865, BGV007931, BGV007989, Brandywine, Floradade, EA00371, EA00990, LYC1410, PAS014479, PI169588 and PI303721, using BLAT analysis19. The Golden Gate assembly approach was used to combine multiple sgRNAs into the destination expression vector20. The primers used for construct generation are listed in Supplementary Table 9 and the final constructs produced are listed in Supplementary Table 1.

Agrobacterium-mediated transformation of tomato

CRISPR−Cas9 constructs were transformed into Agrobacterium strain GV3101 and incubated on YEP agar plates with antibiotic selection (10 µg ml−1 gentamycin, 50 µg ml−1 kanamycin and 20 µg ml−1 rifampicin). PCR amplification was used to validate that transformed colonies contained plasmid. For tomato transformation, true leaves were taken from 4-week-old tomato plants grown in sterile culture and used as explants for transformation according to a previously described method21. Briefly, fresh leaves were cut into approximately 6 × 6 mm2 pieces and incubated in MS-I medium (4.3 g l−1 MS salt including vitamins, 100 mg l−1 myo-inositol, 30 g l−1 saccharose, 7 g l−1 PhytoAgar, 2 mg ml−1 zeatin riboside, 73 mg ml−1 acetosyringon, 0.1 mg ml−1 IAA, pH 5.9) overnight at 25 °C in the dark. Agrobacterium seed cultures were prepared by inoculation of liquid YEP with a single transformed colony, followed by incubation overnight at 28 °C with shaking at 200 rpm. The next day, when the OD600 reached between 0.4–0.6, liquid cultures were diluted 1:20 using liquid lysogeny broth (LB) without antibiotics. Dissected leaves were incubated in Agrobacterium solution for 15 min with occasional moderate shaking, followed by removal of excess liquid and incubation on MS-I medium for 48 h in the dark at 25 °C. Thereafter, leaf pieces were transferred onto MS-II (4.3 g l−1 MS salt including vitamins, 100 mg l−1 myo-inositol, 30 g l−1 saccharose, 7 g l−1 PhytoAgar, 1.5 mg ml−1 zeatin riboside, 100 mg l−1 kanamycin and 500 mg l−1 carbenicillin, pH 5.9) plates and incubated under long-day conditions at 25 °C. The cultivation medium was refreshed every 2 weeks. Emerging shoots were transferred to MS-III rooting medium (2.15 g l−1 MS salt including vitamins, 50 mg l−1 myo-inositol, 15 g l−1 saccharose, 7 g l−1 PhytoAgar, 50 mg l−1 kanamycin and 250 mg l−1 carbenicillin, pH 5.9) and incubated under long-day conditions at 25 °C. Rooted plantlets were screened for transgenesis by PCR (AtCAS9-F and AtCAS9-R; NPT-35S-F and NPT-35S-R) and positive plants were transferred to soil.

Alexander staining and scanning electron microscopy

Alexander staining was performed to determine pollen viability in individual tomato plants22. Mature pollen was collected from mature open flowers using a vibrating tool and stained using a commercial Alexander staining solution. Images were acquired using a Zeiss Axioplan 2 imaging fluorescence microscope equipped with a ZEISS Axiocam 208 color microscope camera, and the images were analyzed using Zeiss Labscope v3.1.

Scanning electron microscopy was performed as previously described23. Pollen grains of mature open flowers were collected into 2-ml Eppendorf tubes and then coated with palladium gold using a Polaron Sputter Coater SC 7600 (Quorum Technologies). The pollen was spread on 25.4-mm specimen mounts (or stubs) (Plano, no. G399F) using 25-mm conductive carbon adhesive tabs (Plano, no. G3348). The results were examined using a field emission scanning electron microscope (Supra 40 VP, Zeiss) with an acceleration voltage of 3 kV.

High-throughput pollen size analysis

The Multisizer 4e (Beckman Coulter) was used to measure the particle diameter and volume of pollen samples derived from single open flowers24. Pollen was collected into 10 ml ISOTON II diluent (Beckman Coulter) in 25-ml Accuvette vials (Beckman Coulter) and measured according to the Multisizer 4e user’s manual using a 100-µm aperture tube with an aperture current of 800 µA. Three different replications with an analytic volume of 1,500 µl were performed. Data were analyzed using Multisizer 4e and GraphPad Prism 9 software, and P values were calculated using standard one-way ANOVA.

Statistics and reproducibility

For single-pollen diameter in wild-type and MiMe plants, we used the Wilcoxon rank-sum test in R (wilcox.test()). Exact P values cannot be reported because of the ties of data points within and between datasets. The following P values were calculated: wild type versus Micro-TomMiMe (<2  10−16); MbTMV-MT F1 versus MbTMV-MTMiMe (<2  10−16); Maxeza F1 versus MaxezaMiMe (<2  10−16); Funtelle F1 FuntelleMiMe (<2  10−16).

For single-fruit weight data of 4-Hap plants, we used ordinary one-way ANOVA followed by Šídák’s multiple-comparison test. For comparison of MbTMV-MT F1 with MbTMV-MT-FuntelleMiMe genotypes, the following exact P values were calculated: MbTMV-MT F1 versus Funtelle F1 (P = 0.004); MbTMV-MT F1 versus MbTMV-MT-FuntelleMiMe-1 (P = 5.54 × 10−5); MbTMV-MT F1 versus MbTMV-MT-FuntelleMiMe-2 (P = 0.0035); MbTMV-MT F1 versus MbTMV-MT-FuntelleMiMe-3 (P = 0.0012); MbTMV-MT F1 versus MbTMV-MT-FuntelleMiMe-5 (P = 2.79 × 10−4). For comparison of Funtelle F1 with MbTMV-MT-FuntelleMiMe genotypes, the following exact P values were calculated: Funtelle F1 versus MbTMV-MT-FuntelleMiMe-1 (P = 0.65); Funtelle F1 versus MbTMV-MT-FuntelleMiMe-2 (P = 0.78); Funtelle F1 versus MbTMV-MT-FuntelleMiMe-3 (P = 0.96); Funtelle F1 versus MbTMV-MT-FuntelleMiMe-5 (P = 0.96). For comparison of MbTMV-MT F1 with MbTMV-MT-MaxezaMiMe genotypes, the following exact P values were calculated: MbTMV-MT F1 versus Maxeza F1 (P = 3.25 × 10−6); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-1 (P = 0.02); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-2 (P = 0.05); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-3 (P = 0.69); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-4 (P = 0.44); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-5 (P = 0.02); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-6 (P = 0.25); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-8 (P = 0.24); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-9 (P = 2.39 × 10−7); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-10 (P = 0.18); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-11 (P = 0.69); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-12 (P = 0.65); MbTMV-MT F1 versus MbTMV-MT-MaxezaMiMe-13 (P = 0.05). For comparison of Maxeza F1 with MbTMV-MT-MaxezaMiMe genotypes, the following exact P values were calculated: Maxeza F1 versus MbTMV-MT-MaxezaMiMe-1 (P = 2.92 × 10−4); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-2 (P = 1.35 × 10−4); Maxeza versus MbTMV-MT-MaxezaMiMe-3 (P = 1.98 × 10−5); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-4 (P = 6.01 × 10−5); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-5 (P = 1.43 × 10−4); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-6 (P = 7.69 × 10−5); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-8 (P = 1.74 × 10−4); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-9 (P = 0.74); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-10 (P = 1.47 × 10−4); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-11 (P = 2.14 × 10−5); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-12 (P = 2.44 × 10−5); Maxeza F1 versus MbTMV-MT-MaxezaMiMe-13 (P = 1.34 × 10−4). No fruit data were available for MbTMV-MT-FuntelleMiMe-4 and MbTMV-MT-MaxezaMiMe-7 because they did not grow to maturity.

Ploidy determination and flow cytometry analysis

Ploidy in plants was determined using flow cytometry. In brief, one piece of fresh young tomato leaf was chopped using a sharp razor blade in 550 µl Galbraith’s buffer (45 mM MgCl2, 30 mM sodium citrate, 20 mM MOPS, 0.1% (v/v) Triton X-100, pH 7.0)25. Next, the slurry was passed through a 30-µm CellTrics green filter (04-0042, Sysmex). Subsequently, 20 µl DAPI (100 µg ml−1) was added to 500 µl of the filtered sample, followed by incubation for 15 min. The CytoFLEX V5-B5-R3 flow cytometer was used following the manufacturer’s instructions. Daily quality control was performed using CytoFLEX Daily QC fluorospheres (B53230). For data collection, 10,000 events per sample were acquired in fast mode for each independent measurement. The final data were analyzed using CytExpert (Beckman Coulter) and FCS Express 7 software.

Chlorophyll content measurements

Leaf chlorophyll contents of control and 4-Hap plants were measured using the AtLEAF tool ( atLEAF CHL values were converted into soil plant analysis development (SPAD) units using a web tool ( and then converted into total chlorophyll content (mg cm−2). For each plant, leaves were counted from the shoot apical meristem toward the ground to identify the fifth leaf (which was in all cases a mature leaf), which was used for analysis. Four different leaflets on each leaf were measured for chlorophyll content and six measurements were performed per leaflet (totaling 24 measurements per genotype).

DNA extraction and sequencing of the CRISPR−Cas9 target sites

Genomic DNA was extracted using a BioSprint 96 DNA Plant Kit (Qiagen). Mutations at CRISPR−Cas9 target sites were analyzed using a combination of PCR amplification, gel electrophoresis and Sanger sequencing or Illumina sequencing. The presence of larger insertion/deletion mutations was first checked by staining agarose gels with Gelgreen (Sigma), imaging and analysis. Sanger sequencing was conducted on PCR product diluted 1:30 using the Mix2Seq Kit (Eurofins Genomics). The produced ABI files were analyzed using the ICE analysis tool (Synthego; Illumina-based amplicon sequencing was modified from a previously described protocol26. M13F (TGTAAAACGACGGCCAGT) and M13R (CAGGAAACAGCTATGAC) were used as bridge sequences on target-specific PCRs and were later used to introduce Golay barcodes (from for each sample in a second round of PCR. After PCR purification, library construction was performed and Illumina sequencing (2 × 150-bp paired end) was done at the Max Planck Genome Centre in Cologne (MPGC Cologne). Sequencing data were analyzed using CLC Main Workbench 21.0.5 software (Qiagen) where the reads were demultiplexed and aligned to the reference (tomato genome, version SL4.0 and annotation ITAG4.). In addition, fixed ploidy variant detection was performed to visualize the specific mutation sites of each gene among different chromosomes. The required variant probability parameter was more than 96% and the coverage and count filter was set to 5% minimum frequency.

Introduction of tomato plants into in vitro culture and embryo rescue

Young side shoots of plants were selected and cut on a super-clean bench. Shoots were sterilized for 15 min in a 1:4 dilution of commercial bleach with 0.02% Tween and then washed three times with Milli-Q. This was followed by transplantation to 0.5 MS-10 medium and subculture on the same medium after 4–5 weeks.

Chromosome spreads

Unopened flower buds (meiotic stages) were collected in 1.5 ml of Carnoy’s fixation buffer (3:1 (v/v) absolute ethanol:acetic acid) and incubated under vacuum for 20 min. The buffer was then refreshed and the material was incubated overnight at room temperature until the tissue turned white. The fixation buffer was then replaced with 75% ethanol and samples were stored at 4 °C. Chromosome spreads were performed as described in a previous protocol27,28 with the following modifications: the individual anthers were separated (the length of anthers ranges from 1.5 to 3 mm in meiotic stages) from the flower buds and digested with enzyme solution (0.3% (w/v) cellulase RS, 0.3% (w/v) pectolyase Y23, 0.3% (w/v) cytohelicase in citrate buffer, pH 4.7) for 2 h at 37 °C. Meiocytes were released from two or three anthers in 45% acetic acid by crushing the anthers using tweezers, followed by covering them with a coverslip, taking care to avoid bubbles. The slides were immersed in liquid nitrogen until no sound was heard and then the coverslip was removed. The slides were successively dried by applying 70%, 85% and 100% ethanol (5 min per concentration). Finally, 6 µl DAPI (10 µg ml−1) was added to stain the chromosomes after the slides had dried. Images of chromosomal spreads were acquired using a Zeiss Axio Imager Z2 upright microscope and analyzed using ZEN blue software (Zeiss).

Homologous protein sequence identification, characterization and phylogenetic tree analysis

The protein sequences of AtOSD1 (AT3G57860), AtCYCLIN A1;2 (AT1G77390), AtSPO11-1 (AT3G13170) and AtREC8/SYN1 (AT5G05490) were acquired from the Arabidopsis database TAIR (The Arabidopsis Information Resource; and then BLAST was performed against the phytozome database (, the UniProt protein database (, the Solanaceae Genomics Network database ( and the NCBI database ( to identify homologous protein sequences in other species. Proteins were aligned using Clustal X2 followed by the construction of a phylogenetic tree using MEGA11. Gene structure images were created using the Exon-Intron Graphic Maker (

Whole-genome sequencing of parental and offspring samples

High molecular weight DNA of Micro-Tom, Funtelle and Maxeza was isolated from 1.5 g of young leaf material using a NucleoBond HMW DNA kit (Macherey Nagel). DNA quality was assessed using a FEMTOpulse device (Agilent) and DNA quantity was measured using a Quantus Fluorometer (Promega). High-fidelity (HiFi) libraries were prepared according to the manual “Procedure & Checklist—Preparing HiFi SMRTbell Libraries using SMRTbell Express Template Prep Kit 2.0” with initial DNA fragmentation using a Megaruptor 3 (Diagenode) and final library size binning into defined fractions using SageELF (Sage Science). The size distribution was again controlled by FEMTOpulse (Agilent). Two size-selected libraries were sequenced per genotype on single SMRT cells (that is, a total of six SMRT cells) on a Pacific Biosciences Sequel II or Sequel IIe device at the MPGC Cologne with binding kit 2.0 and Sequel II Sequencing Kit 2.0 for 30 h (Pacific Biosciences).

For Micro-Tom, a chromatin conformation capture library was prepared using 0.5 g of young leaf material as the input. All treatments were performed according to the recommendations of the kit vendor (Omni-C, Dovetail) for plants. As a final step, Illumina-compatible libraries were prepared (Dovetail) and test-sequenced (2 × 150 bp paired end) on an Illumina NextSeq 2000 device at the MPGC Cologne, followed by sequencing on a NovaSeq 6000 at Novogene for increased coverage. In addition, Micro-Tom genomic DNA was extracted using the Qiagen DNeasy Plant Mini Kit and a BGI Plug-in Adapter Kit was used to prepare a sequencing library that was subsequently sequenced at BGI.

MbTMV, Funtelle and Maxeza genomic DNA was extracted using the DNeasy Plant Mini Kit (Qiagen). A PCR-free sequencing library was prepared at BGI and subsequently sequenced at BGI. MbTMV-MT F1 (6 samples), MbTMV-MT F2 (6 samples), MbTMV-MTMiMe-A offspring (6 samples), MbTMV-MTMiMe-B offspring (6 samples), Funtelle F1 (2 samples), Maxeza F1 (2 samples), MbTMV-MTMiMe MaxezaMiMe offspring (13 samples) and MbTMV-MTMiMe FuntelleMiMe offspring (5 samples) were profiled by Illumina sequencing. Briefly, small amounts (around 1 cm × 1 cm) of leaf samples were collected into 96-well plates and then the DNA was isolated at the MPGC Cologne using a NucleoMag Plant Kit (Macherey Nagel, 744400.4) to extract DNA on a robotic device (KingFisher, Thermo) followed by TPase-based DNA library preparation. The pooled libraries were sequenced using an Illumina NovaSeq 6000 machine at Novogene. Raw read and mapped read numbers are provided in Supplementary Table 7.

Genome assemblies

For Micro-Tom, Hifiasm v0.16.1-r375 (ref. 29) was used with option -l0 to assemble the HiFi reads. First, omni-C (Dovetail) paired-end reads were mapped separately using Burrows−Wheeler Aligner v0.7 (ref. 30), followed by the addition of read mate scores, sorting and removal of duplicate reads using Samtools v1.9 (ref. 31). The resulting bam file was converted to a bed file and sorted (-k 4) using Bedtools v2.30 (ref. 32). Subsequently, one single round of Salsa v2.2 (ref. 33) was performed using the following optional settings: -e DNASE -m yes -p yes. A modified version of the script was used to convert the Salsa2 output to a Hi-C file, which was used within a local installation of Juicebox ( v1.11.08 to generate Hi-C contact plots. Assemblies scaffolded with the automated Salsa2 pipeline were further fine-tuned, with unplaced smaller scaffolds manually placed within larger scaffolds, based on Hi-C interaction and alignment to the MbTMV genome9.

Haplotype-resolved assemblies of Funtelle and Maxeza F1 hybrids were generated by running Hifiasm v0.16.1-r375 (ref. 29) on HiFi reads with default settings. Alignment of raw contigs against the MbTMV reference genome9 was generated using minimap2 v2.24-r1122 (ref. 34) and visualized using D-GENIES v1.4.0. For each haplotype, contigs were anchored and scaffolded to chromosome-scale pseudomolecules using Ragtag v2.1.0 (ref. 35). K-mer analysis and genome size estimation were performed using genomescope v1.0 (ref. 36).

Marker detection

We first identified and assigned SNP markers into two haplotypes. To detect markers between MbTMV and Micro-Tom, we first aligned the short reads (BGI) and long reads (HiFi) of both parental genomes against MbTMV9 using bwa-mem v0.7.17 (ref. 37) and minimap2 v2.24-r1122 (ref. 29), respectively. SNPs were then detected using GATK HaplotypeCaller v4.2.4.1 and hard-filtering38. We retrieved homozygous Micro-Tom SNPs detected using both HiFi and Illumina reads that did not match any SNPs detected during MbTMV data alignment, ensuring that no ambiguous markers remained. We also selected markers that differed between the MbTMV-MT F1 hybrid and the Funtelle F1 hybrid by selecting homozygous Funtelle SNPs that did not overlap with any MbTMV or Micro-Tom SNPs. A similar approach was used to detect segregating markers between the MbTMV-MT F1 hybrid and the Maxeza F1 hybrid.

To identify unique markers in each of the four different haplotypes in the tetraploid tomato plants made by polyploid genome design (MbTMV-MTMiMe MaxezaMiMe offspring; MbTMV-MTMiMe FuntelleMiMe offspring), we initially needed to discover markers that differed between Funtelle haplotype 1 and Funtelle haplotype 2 and, separately, markers that differed between Maxeza haplotype 1 and Maxeza haplotype 2. In both cases, the same method was used, but for simplicity, the Funtelle case is described below. Both scaffolded assemblies of Funtelle haplotypes were aligned against MbTMV using minimap2 v2.24-r1122 and were then processed using MUMmer dnadiff39 to identify SNPs. To confirm the SNPs were heterozygous between the two Funtelle haplotypes, we retrieved assembly-based SNPs that matched heterozygous SNPs reported from Funtelle HiFi and BGI reads. We retrieved those that did not overlap with any SNPs from MbTMV and Micro-Tom (Supplementary Fig. 25a). The resulting set of SNPs uniquely identified Funtelle haplotypes in both MbTMV and Micro-Tom. For unique Micro-Tom SNPs, we retrieved homozygous SNPs that did not overlap with any MbTMV or Funtelle SNPs. For unique MbTMV SNPs, we reported the genomic positions that showed homozygous SNPs for both Micro-Tom and Funtelle. To reduce false markers, we excluded a region on chromosome 9 between 5 Mb and 58 Mb, which is part of an introgression in MbTMV from the wild tomato relative S. peruvianum9. The resulting set of SNPs is considered to include unique parental markers.

Determination of crossovers and aneuploidy

Trimming and quality checks of Illumina reads were done using TrimGalore ( The alignment of reads and SNP calling against the MbTMV genome were performed using bwa-mem v0.7.17 and GATK HaplotypeCaller v4.2.4.1. For each sample, we computed the average allele frequency in a sliding 1-Mb window with a step size of 50 kb. We visually inspected deviations from the expected frequency to differentiate F2 from F1 and MiMe samples and to infer the presence or absence of recombination. Some of the samples showed unexpected allele frequencies in some genomic regions or chromosomes; therefore, we developed a script to identify regions with deviations in both allele frequency and average read coverage to infer chromosome fragmentation/aneuploidy. Finally, to determine whether all four haplotypes were present in the tetraploid tomato plants (MbTMV-MTMiMe MaxezaMiMe offspring; MbTMV-MTMiMe FuntelleMiMe offspring), we counted the number of unique parental markers that could be observed in each sample and compared this with the number in the control hybrids (MbTMV-MT F1, Funtelle F1 and Maxeza F1).

Predicting gene dosage and genotyping

We selected specific genes/regions of agronomic interest (tobacco mosaic virus resistance (Tm-22, Solyc09g018220)40, M. incognita resistance (Mi, Solyc06g008720)41,42, self-pruning (SP, Solyc06g074350)43, dwarf (D, Solyc02g089160)44 and Fusarium wilt resistance (I, Solyc11g011180)19 to count alleles and dosage. For genes with known causative mutations, we checked both the assemblies and BGI short reads. SyRI v1.6.3 was used to detect structural variations relative to the MbTMV reference and to find haplotypes with introgressions45. For the identified genomic rearrangements, genomic variations among the haplotypes, including syntenic regions, inversions, translocations and duplications, were highlighted in gray, yellow, light green and blue, respectively. SNP density was also computed to infer regions of wild introgression using ONT data from Fla.8111B and LA1589 to check for the presence of resistance gene I19. To directly genotype the same genes in each 4-Hap plant, we used either the known causative mutation or the haplotype of the resistant accession using GATK HaplotypeCaller v4.2.4.1. To annotate the assemblies and count the number of non-Lycopersicum genes, CDS sequences from S. peruvianum46 were mapped to each of our parental assembles using minimap2. We identified wild-type genes within the large S. peruvianum introgressions on chromosomes 6 and 9.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.